Backed Research Agent

Backed Research Agent is an analyst-grade web3 diligence engine for venture-style underwriting. It transforms heterogeneous source material (websites and documents) into a structured investment output for IC workflows, risk teams, and dashboard systems.

Scope

The engine is designed for early-stage to growth-stage crypto/web3 opportunities where information quality is inconsistent and downside protection is critical.

Core capabilities

1. Ingestion

• URL fetching with retry/backoff and fallback strategy for 429/5xx surfaces.
• Local file ingestion (txt, md, csv, json, pdf, docx).

2. Deterministic underwriting

• Multi-dimension scoring with explicit weights and confidence.
• Positive/negative evidence accounting for each analytical dimension.

3. Risk and team evaluation

• Structured risk register (probability, impact, severity, mitigation metadata).
• Team underwriting across founder visibility, execution history, domain fit, hiring signal, governance trust.

4. Market and fundraising context

• Live market signals (BTC/ETH, dominance, TVL, sentiment, regime).
• Raise signal extraction and benchmark comparison with raising difficulty score.

5. Decision output and dashboard payload

• Executive summary and recommendation.
• Project assessment and investment memo.
• Dashboard-native section with KPIs, alerts, gates, and dimension rows.

Runtime modes

• rules: deterministic engine only.
• auto: deterministic engine plus AI refinement when OPENAI_API_KEY is available.
• ai: AI refinement required.

AI refinement improves narrative and advanced analytical fields but does not replace deterministic pipeline structure.

Installation

pip install -e .

Configuration

Create .env from template.

Unix/macOS:

cp .env.example .env

Windows PowerShell:

Copy-Item .env.example .env

Supported variables:

• OPENAI_API_KEY (optional)
• BACKED_API_KEY (default: GIT3PRIVATE)
• BACKED_API_HOST (default: 127.0.0.1)
• BACKED_API_PORT (default: 8000)
• BACKED_API_CORS_ORIGINS (default: *)

CLI usage

Primary command:

backed-research-agent --source "https://project-site.com" --json --output "./out/report.json"

Compatibility alias:

investment-agent --source "https://project-site.com" --json --output "./out/report.json"

Module entrypoint:

python -m investment_agent.cli --source "https://project-site.com" --json --output "./out/report.json"

CLI options:

• --source repeatable URL or local file input
• --mode auto | ai | rules
• --model LLM model for AI mode (default gpt-4.1-mini)
• --json print full JSON output
• --output write full JSON output to file

API usage

Start API service:

backed-research-agent-api

Available endpoints:

• GET /v1/health
• POST /v1/analyze (API key required)

API authentication:

• X-API-Key: <key>
• or Authorization: Bearer <key>

Request payload supports:

• source (single source)
• sources (multiple sources)
• mode
• model

Response includes:

• request metadata (request_id, sources, mode, model)
• full underwriting result under result

Output contract

Top-level analysis payload sections:

• score
• executive_summary
• recommendation
• project_profile
• project_assessment
• risk_register
• team_assessment
• market_snapshot
• fundraising_context
• dashboard
• findings
• sources

Schema reference: docs/OUTPUT_SCHEMA.md

Technical documentation

• Integration.md
• docs/API.md
• docs/ARCHITECTURE.md
• docs/VC_FRAMEWORK.md
• docs/DASHBOARD_INTEGRATION.md
• docs/RESEARCH_BASIS.md
• docs/ROADMAP.md
• docs/PROTOCOL_INTEGRATIONS.md

Near-term integration roadmap

Planned protocol roadmap includes x402 on MegaETH as a payment/authentication primitive for premium research execution paths. This integration is planned and not active in the current release.

Development

python -m unittest discover -s tests -v

Disclaimer

Backed Research Agent is a decision-support system and not financial, legal, or investment advice.